Abstract: In this paper we propose an efficient method for estimating seemingly unrelated multivariate regression models. The gain in efficiency can be partially assessed by Hotelling's canonical correlations . We apply this method to the estimation problem and the concomitants selection problem in growth curve models.
Key words and phrases: Seemingly unrelated multivariate regression models, gain in efficiency, Wishart distribution, seemingly unrelated growth curve models, concomitants selection, two-stage estimate.